Kalles Fraktaler 2 + GMP

  • 232 Replies
  • 11667 Views

0 Members and 1 Guest are viewing this topic.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #210 on: June 04, 2018, 07:24:54 PM »
kf-2.12.13.1 released! get it from https://mathr.co.uk/kf/kf.html#download

  -  fix severe performance regression in previous version
  -  upgrade to Boost 1.67.0
  -  upgrade to MPFR 4.0.1-p6

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2161
« Reply #211 on: June 06, 2018, 05:14:06 AM »
It's really fast compared to 2.13, more than a factor 2. I wonder why 2.13 is so slow. Just adding the update for the derivative should slow things down by no more than a factor 2, even assuming no overhead (ref. orbit etc.), no? (Referring to normal M-set here.)

Even if you render with NDE fast using 2.12, 2.13 is still useful to get a low resolution preview for coloring with ADE. NDE colors change drastically with increased resolution so your hi-res image may not turn out the way you expected from a lo-res preview. ADE in 2.13 is close to a (very) hi-res coloring with NDE.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #212 on: July 25, 2018, 08:15:29 PM »
A single dot appears in the middle of a mini.

I think I fixed this locally in my 2.13 branch.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #213 on: August 14, 2018, 02:49:21 AM »
https://mathr.co.uk/kf/kf.html#kf-2.12.14 new release of the stable branch, minor bugfixes

  -  fix for "a single dot appears in the middle of a mini" (reported by gerrit)
  -  possible fix for "iterations reset low when zooming fast"
  -  upgrade to GSL 2.5
  -  update documentation

Offline CFJH

  • *
  • Fractal Frankfurter
  • *
  • Posts: 506
« Reply #214 on: September 20, 2018, 11:03:54 PM »
Hello,

currently I've an ugly problem with kf 2.12.14 and the exif information stored in a jpg image.
When the exif information reaches multiple of 64Kb in size, some control characters are inserted. See attachment 3.
I'm parsing the exif information (with PHP, here "exif_read_data()" is used) to generate the zoom factor label in the final video. But this fails now.
Also the exiftool from Linux can't get the complete exif information, it stopps at the first misplaced character (attachment 1).

test.jpg is the file itself (not finished, shows the standard mandelbrot)
kf 2.13.x doesn't have this problem.

How can I handle this problem?
Does kf handle the problem correcty or are all images broken now ?

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #215 on: September 21, 2018, 02:18:04 PM »
Does the JPEG metadata load correctly into KF?

The splitting into multiple comment chunks is due to a limitation of the JPEG comment format:

JPEG comments start with 4 bytes:
0xFF 0xFE L1 L2
the length of the comment that follows is ((L1 << 8) | L2) - 2
being only 2 bytes, that limits the length of each chunk to 64KiB - 3B

KF should concatenate the chunks when reading, but maybe there is a bug in this process?

Not all software handles multiple comment chunks, so this may be a bug in that code, but I don't know if multiple comment chunks are strictly allowed by the JPEG standard, in which case KF is doing it all wrong. Use PNG (slow) or TIFF (huge) to avoid this issue.

Here is the source code for KF's JPEG comment loading, should be fairly easy to port to PHP or other languages (but check that KF loads it correctly, first!):
https://code.mathr.co.uk/kalles-fraktaler-2/blob/kf-2.12.14:/fraktal_sft/jpeg.cpp#l244

Offline CFJH

  • *
  • Fractal Frankfurter
  • *
  • Posts: 506
« Reply #216 on: September 21, 2018, 11:36:01 PM »
Hi,

I've tested it with kf 2.12.14.  Loading a file with zoom 1.66E178929 (6 chunks in exif data) and saved again (without writing to finish the calculation) to a new file shows that all data was read (no differences in the exif data).
kf 1.13 does this in the same way (my statement in the first post was wrong, havn't checked it correctly).

In the meantime I've adjusted my php script by reading the exif data manually, now it is working again.  :)

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #217 on: October 29, 2018, 08:44:56 PM »
kf-2.12.15 released https://mathr.co.uk/kf/kf.html#kf-2.12.15 bugfix release for the oldstable branch

  -  bugfix: corrupt image at transition between number types (eg near e600) (reported by CFJH)
  -  bugfix: changing "threads per CPU" during rendering crashes (reported by CFJH) (the menu is now disabled during rendering)
  -  bugfix: reading PNG metadata works even if it is moved after the image data chunks and has a miscapitalized "Comment" keyword
  -  bugfix: ghc-8.6 compatibility (MonadFail)
  -  bugfix: translation with reuse reference (reported by Dinkydau)
  -  bugfix: don't add secondary references when auto solve glitches is disabled
  -  upgrade to Boost 1.68.0
  -  upgrade to MPFR 4.0.1-p13
  -  upgrade to libpng 1.6.35

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #218 on: March 20, 2019, 05:30:29 AM »
kf-2.12.16 released https://mathr.co.uk/kf/kf.html#kf-2.12.16 bugfix release for the oldstable branch

 -   bugfix: malformed location no longer crashes (reported by Foxxie)
 -   upgrade to mpfr 4.0.2
 -   upgrade to glm 0.9.9.4
 -   upgrade to libpng 1.6.36
 -   upgrade to libtiff 4.0.10
 -   upgrade to boost 1.69.0

Offline Fluoroantimonic_Acid

  • *
  • Fractal Fanatic
  • ***
  • Posts: 22
« Reply #219 on: May 01, 2019, 03:12:42 AM »
Hey, So I've been using the latest version in your most stable branch, 2.12.16 for a while now, and recently I've came across what I think may be a bug in some really extreme locations.

What happens is that in areas nearing 1 billion iterations, the program tends to not automatically update the screen, and a window resize is almost always needed in order to see what the program is doing. What's more, is that the progress counter at the bottom left stops after the A%, and sometimes only when the R% value is at it's max.

I'm not sure if this is a normal course of action for the program, or if it is indeed a bug, but in any case, I thought it would be important to notify you on the issue.

Below I've attached the current location I'm working on in case if you really want to take a closer look at it.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #220 on: May 01, 2019, 03:01:09 PM »
There are probably all kinds of issues when the iteration count gets so large, because the number type used is 32bit signed, which means if you go above 2^31-1 (about 2 billion) then undefined behaviour results (it could go negative, for example).  There was at least one bug triggered by the calculation of the % for display overflowing, but I think I fixed that.

Fixing KF to support such high iteration counts (by switching to 64bit int for iteration counts) would be a massive undertaking, and it might be hard to guarantee that all possible bugs in that area have been fixed.  It would also use up to 50% more memory (and disk space for KFB), which might be an issue for large renders at non-extreme iteration counts.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #221 on: May 01, 2019, 05:20:59 PM »
Hm.  I think overflow is not the problem here. I can reproduce some issues:

1. Sometimes the status timer etc doesn't get going, just says Done.  Pressing ESC followed by F5 usually fixes this, restarting rendering in the process.  Probably a race condition or other logic error.

2. The pixels are not updated/displayed soon enough, so the image remains blank / showing previous contents.  Pressing ESC aborts rendering and updates the display, but obviously this is undesirable.  See screenshot attached, after pressing ESC.  There is a balance between updating pixels immediately, and keeping the old (scaled) image for interactive nagivation.  Maybe a timeout should start updating pixels after 5secs or so, not sure why it doesn't update at the moment...

Offline Fluoroantimonic_Acid

  • *
  • Fractal Fanatic
  • ***
  • Posts: 22
« Reply #222 on: May 01, 2019, 10:26:52 PM »
Do you think you may be able to get some failsafe procedure like that incorporated into the program eventually?
 
Although it isn't a big issue for me currently, I do plan on making a zoom on that location when I can get it to the minibrot.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #223 on: May 02, 2019, 10:13:52 PM »
It's certainly theoretically possible to do it.  Whether I will be motivated to do it in the foreseeable future is another thing entirely.

It is a bit lonely working on KF without any other developers helping, and I don't use it so much myself (I use my own mandelbrot-perturbator-gtk for low-to-medium-depth quadratic Mandelbrot set research due to its annotation features like drawing external rays and marking periods, and I use my own still-experimental et-gtk for exploring other formulas including hybrids as it has a formula compiler).

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #224 on: May 02, 2019, 11:39:32 PM »
2. The pixels are not updated/displayed soon enough, so the image remains blank / showing previous contents.

This is now fixed in the source code repository.  Before it was using int for the percentage of completion, and testing that it was >0, so it would only refresh the image after 1% of the pixels were completed.  I changed it to use double, so now it starts refreshing the image as soon as the first pixel is completed.

Will be in the next releases of 2.12/2.13/2.14.  Not sure when that will be, let me know if there are any other bugs that need fixing.

Thanks for the bug report!


xx
Kalles Fraktaler 2.13

Started by claude on Kalles Fraktaler

90 Replies
4661 Views
Last post March 20, 2019, 05:32:26 AM
by claude
xx
Kalles Fraktaler 2.14

Started by claude on Kalles Fraktaler

233 Replies
7386 Views
Last post July 07, 2020, 05:59:03 AM
by gerrit
xx
Another version of Kalles Fraktaler?

Started by greentexas on Kalles Fraktaler

9 Replies
1099 Views
Last post September 22, 2017, 02:59:16 PM
by greentexas
xx
Feature ideas for Kalles Fraktaler

Started by FractalAlex on Kalles Fraktaler

11 Replies
179 Views
Last post June 05, 2020, 04:01:18 PM
by FractalAlex
xx
Kalles Fraktaler Movie Maker64

Started by Bill Snowzell on Fractal movie gallery

0 Replies
297 Views
Last post February 01, 2018, 09:29:04 PM
by Bill Snowzell